Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.
Engineering & simulationNoteworthy AI computer vision models
The AI was trained on diverse environmental data (weather, lighting, geography) to generalize across different pole types and conditions, improving defect identification accuracy.
“So what the broad deployment over the years has done for us is given us the opportunity to get all these data samples and funnel that into iterative model training to make it generalize really well for any type of condition or any type of poll composition, as we call it.”
ManufacturingNoteworthy AI computer vision modelsevaluation of 10% of grid per year to full grid coverage
The AI autonomously identifies grid defects (leaning poles, damaged components) during routine vehicle operations, shifting the workflow from manual inspection to automated flagging for repair crews.
“So, you know, I explained before how the status quo is, binoculars and notepads driving around very, very labor intensive, right? And because it's very labor intensive, it's very costly, that means historically, it doesn't scale. So if you're the average large utility in the United States, you have quote unquote eyes on. You're evaluating about 10% of your grid a year.”
Logistics & fulfilmentNoteworthy AI edge AIresults in minutes vs 24-48 hours
The AI processes storm damage data at the edge and sends results to the command center in minutes, enabling faster mobilization of crews and inventory compared to the previous 24-48 hour manual process.
“With a technology like ours, enabled by IOT and real-time AI in the edge for these storm specific damages that we've developed computer vision models for, we can send the results up to the command center in minutes on a purple basis.”